Jump to content

Any way to make a Windows 7 partition with larger than 4k clusters?


Recommended Posts

...That is, the partition where the BCD resides. I have a specific setup which I don't want to disturb:

HnWPIW7.png

This is as seen from Windows 7, which is on the second partition (C:). Dual boot is with XP, which is on the first partition, seen here as F: . When XP is booted, letters are reversed (XP, first partition, is C: and Win7, the second, is F:). It works perfectly for me because every OS is "isolated" in its own partition, sees itself as drive letter C: and the BCD is on the Win7's partition. I use custom boot selector/loader (OSL2000) which is the only thing (tiny) inserted into the first sector (or boot sector, or whatever it's called) of the disk.

Now, this is an SSD and I want to make the cluster size equal to the NAND page size, which is 8k for this Samsung 840 Pro. The purpose is longevity of the SSD. In many places I have read that this is impossible for Windows 7 (BCD). Is that still so? I don't want to create another partition.

I have already enlarged clusters (non-destructively) of the other two partitions from 4k to 8k with Paragon Hard Disk Manager (bootable CD), but when I do it to the Win7 partition, it doesn't boot. Even Paragon says so in the help file. Luckily it was reversable, but of course I still have backups.

So, can I do anything? This is the most written to partition here.

*Edit: and by the way, all partitions are aligned, and I think the alignment won't be disturbed if I don't resize them. Partitions start at offset:

1048576
16013852672
49568284672

GL

Edited by GrofLuigi
Link to comment
Share on other sites


I know very little about this, so I did some search. Does this help any?

I can confirm for you that Windows does not support sector sizes greater than the Memory Manager's page size.

Its from here: http://www.osronline.com/showthread.cfm?link=235324

The only thing on MS about 8k I can find is refering to DOS.

http://support.microsoft.com/kb/67321/en-us

I checked that since in that forum thread there is a link to a 4k OS support article, but makes no mention of 8k.

Link to comment
Share on other sites

@Trip

Yep :) but that is about Physical Sector size (that has nothing actually to do with cluster size).

@Grofluigi

There may be issues with the boot code (and just with it).

Since you do have some 23 Gb free on that device, you can experiment, making a (temporary) partition using that space.

Care to share the source of your opinion of using an 8 Kb cluster being *any* "better"? :unsure:

jaclaz

Link to comment
Share on other sites

@Grofluigi

There may be issues with the boot code (and just with it).

Since you do have some 23 Gb free on that device, you can experiment, making a (temporary) partition using that space.

Care to share the source of your opinion of using an 8 Kb cluster being *any* "better"? :unsure:

jaclaz

That is some magic free space a Magician from Samsung created for me. :w00t: Or, in other (big) words, over-provisioning is supposed to reduce write amplification and ensure write consistency. I don't care much about that second part (write consistency), but since the manufacturer recommends overprovisioning and I can afford to lose some space (I have enough), I go with it.

If I were to make another (boot?) partition, I'd make it at the beginning of the disk, but I don't want to (for now) since everything is so neat and self-contained.

I would like to have 8 Kb clusters because of "write amplification" or, in this case "erase amplification?". Samsung 840 Pro (and every SSD) erases bytes in pages, which in this case is 8 Kb. So in theory, for a file of under 4 Kb in size, it would unnecessarily erase another block of 4 Kb or leave this block "dirty". Or the garbage collection routine would transfer the contents to another block and then erase these two 4 Kb blocks.

GL

Link to comment
Share on other sites

I would like to have 8 Kb clusters because of "write amplification" or, in this case "erase amplification?". Samsung 840 Pro (and every SSD) erases bytes in pages, which in this case is 8 Kb. So in theory, for a file of under 4 Kb in size, it would unnecessarily erase another block of 4 Kb or leave this block "dirty". Or the garbage collection routine would transfer the contents to another block and then erase these two 4 Kb blocks.

I understand that, I was not-so-sure about that particular SSD having 8 Kb pages (and operating on them as "full pages").

In any case - and with all due respect :) - it seems to me like the idea makes little sense. :w00t::ph34r:

Let's take two extremes for the sake of reasoning.

  1. only 1 (one) file below 4 Kb is in the filesystem
  2. several thousands files below 4 Kb are in the filessystem

If #1 the impact of *whatever* is null.

If #2 while the impact of the *whatever* may be relevant or noticeable, in order to reduce it you will further reduce the available capacity of the SSD considerably.

My suggestion was however to try making a new (smallish partition) (Yes, a "boot" partition) not entirely unlike the one Windows 7 makes on a non-partitioned disk, with just the BOOTMGR and the \boot\BCD and see if you can boot from that one the "main" partition with the 8 Kb sized clusters

.

Since it is just a test you can use a part of the space you have free (it is not like having 100 Mb less than the current 23.85 Gb will change *anything*).

jaclaz

Link to comment
Share on other sites

New partition will be my last resort...

I was asking more if there are any new, or old, or unknown to me (ungoogled by me) :) methods to "repair" a BCD on non-4Kb partitions. Maybe third-party tools have progressed (I've heard of EasyBCD, and i think there are others), or something might be possible to be hex-edited, but I need to know what...

And the reason makes perfect sense (at least in my head, obviously :w00t:) when one remembers that SSDs have finite number of writtes.

GL

I would like to have 8 Kb clusters because of "write amplification" or, in this case "erase amplification?". Samsung 840 Pro (and every SSD) erases bytes in pages, which in this case is 8 Kb. So in theory, for a file of under 4 Kb in size, it would unnecessarily erase another block of 4 Kb or leave this block "dirty". Or the garbage collection routine would transfer the contents to another block and then erase these two 4 Kb blocks.

I understand that, I was not-so-sure about that particular SSD having 8 Kb pages (and operating on them as "full pages").

In any case - and with all due respect :) - it seems to me like the idea makes little sense. :w00t::ph34r:

Let's take two extremes for the sake of reasoning.

  1. only 1 (one) file below 4 Kb is in the filesystem
  2. several thousands files below 4 Kb are in the filessystem

If #1 the impact of *whatever* is null.

If #2 while the impact of the *whatever* may be relevant or noticeable, in order to reduce it you will further reduce the available capacity of the SSD considerably.

My suggestion was however to try making a new (smallish partition) (Yes, a "boot" partition) not entirely unlike the one Windows 7 makes on a non-partitioned disk, with just the BOOTMGR and the \boot\BCD and see if you can boot from that one the "main" partition with the 8 Kb sized clusters

.

Since it is just a test you can use a part of the space you have free (it is not like having 100 Mb less than the current 23.85 Gb will change *anything*).

jaclaz

Link to comment
Share on other sites

I cannot believe that - IF the issue is in the booting only - it can be solved by modifying the \boot\BCD (which AFAIK contains NOT any particular info related to the filesystem and it's cluster size).

If you don't want to add a partition, not even temprorarily for the sake of the test) , you can do with a boot floppy image:

http://www.multibooters.co.uk/floppy.html

an image like described can be mapped and chainloaded by grub4dos alright, and it will need not any re-partitioning, while removing the doubt that it is a BOOTMGR issue of some kind, which is a mere (though educated ;)) guess :ph34r: based on your "it doesn't boot." report (which is not an actual description of what actually happens :whistle:).

About the "sense", what I was trying to highlight is that the theoretical "increased wear" may happen only on a limited number of cells, since the number of files less than 4 Kb that are often re-written on a "normal" Windows 7 system are very few, maybe the relevance of the issue is "amplified" by any file which is a multiple of 4 Kb but not of 8 Kb? :unsure:

And, call me reckless if you want :w00t:, according to these:

http://www.anandtech.com/show/7173/samsung-ssd-840-evo-review-120gb-250gb-500gb-750gb-1tb-models-tested

http://www.anandtech.com/show/6328/samsung-ssd-840-pro-256gb-review

http://www.anandtech.com/show/6459/samsung-ssd-840-testing-the-endurance-of-tlc-nand

one of those thingies are guaranteed for 5 years and are estimated at around 11 years of duration.

By that time we will likely be using Windows 13.4 SP2¼ and 256 Gb ;) will be the space needed for just the bootmanager and configuration file.

jaclaz

P.S.: After all my guess was educated enough:

http://www.terabyteunlimited.com/kb/article.php?id=392

http://support.microsoft.com/kb/2272294

Edited by jaclaz
Link to comment
Share on other sites

"it doesn't boot." report - I forgot the exact words, they were on a black screen, not terribly scary, but not big friendly letters either. ;)

Due to lack of free time, I will suspend my efforts for now and come back to it later.

Thank you Jaclaz

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...